Asking the Right Questions: Benchmarking Large Language Models in the Development of Clinical Consultation Templates